What's in a generation or two?

« previous post | next post »

In a recent post ("Creole birdsong?", 5/9/2008), I mentioned Derek Bickerton's "language bioprogram hypothesis". Derek responded with a long comment. Since I responded to his comment with another post ("A multi-generational bioprogram? Derek Bickerton objects", 5/10/2008), I invited him to respond in kind. The guest post below is the result.

[Guest post by Derek Bickerton]

Frankly I'm at a loss to understand why it makers no difference whether language (birdsong, whatever) is acquired in one generation or several, as a consensus of first responders seem to be claiming. First of all let's put aside the birds. What they do has no necessary connection with what humans do, and this is after all Language Log not Twitter'n'tweet.com.

If there is any kind of biological program for language (and later I'll suggest the most plausible form this might take) there's no logical reason why it should not kick in for the first generation that needs it. Anything multigenerational suggests what Mark explicitly claims, a socializing back-and-forth between people (or birds) in which variants gradually get leveled until you get homogeneity (or something close).

This would be fine if the only problem was pidgin variability, if regularization was all the first creole generation had to do. But that's less than half the story. A pre-creole pidgin isn't just variable, it's seriously impoverished in a way that makes Chomsky's "poverty of the stimulus" look like Donald Trump on a spending spree. It has no complex sentences. It has virtually no function words—the few that make it are used in only a fraction of the contexts that require them, and quite unpredictably at that. Why do you suppose creoles have to bleach lexical items in order to get so many of their function words? Where can the unprecedented structures come from if not from the organisms that are producing them? (Yes, I know there's a substrate and a superstrate, but if anyone wants to argue for them, go ahead, make my day.)

People get all ballsed up over innateness, so here's what I now think. There is no language organ. You don't need one. What happened to humans was they got words (for how, see my next book, Adam's Tongue, now in press). So after all Terry Deacon was right that words not syntax are the most crucial ingredient in language. Once the brain had them, it was going to put them together somehow—that's what brains were built to do, put stuff together, whether it's all the bits and pieces that make up what you see or the sequence of movements that enables you to open wine bottles. How they do this you could probably find out if you raised kids with just one-word-at-a-time input, no examples of even the simplest syntactic structures—I seriously suggest this in Bastard Tongues, step up to the plate any dot-com millionaire who'd like instant notoriety for funding it. Short of that, plantation creoles come as close as Nature's own experiments are ever likely to take us.

You see, what happens in language acquisition is simply this. The way the brain does it causes the child automatically to put words together in certain ways. Sometimes those ways are the right ways, in which the kids are home and dry with no more ado. More often, probably, they're the wrong way, so the child adjusts. There's none of this LAD-using, parameter-setting, rule-inducing Little Linguist stuff. It's the same all over—the child has no way of knowing whether its input consists of a natural language with thousands of years of history or a new pidgin virtually devoid of structure. It just forges ahead regardless, learning more and more words and what they're for (including functions of grammatical items, naturally) and if there's relatively little counterevidence to the child's initial assumptions, what you get is your classic creole.

So there you have it—a simple, logical, unified explanation of language acquisition, language evolution, and language creation. Too simple to be true? Well, I'd love nothing more than for some of these people who say that the bioprogram hypothesis "no longer commands the attention it once did" or flat-out state it's been decisively disproven to strut their best stuff in these columns and we'll see what happens

[Guest post by Derek Bickerton]



8 Comments

  1. John Roth said,

    May 12, 2008 @ 10:32 am

    I find the "single generation" hypothesis unlikely. The reason is simply that language acquisition in children is a multi-stage process where specific kinds of linguistic phenomena are acquired in a fixed order (but not always at the same pace).

    The reason is most likely brain maturation, where the various areas and connections are myelinized at different times.

    Looked at this way, it seems likely that if the cohort of children has to invent its way through one or more stages it won't have the necessary prerequisites for the next stages when the brain is ready to learn them.

    I would submit that this model supports multi-generational creation of a richer and richer creole, ending in a fully functional language.

    John Roth

  2. Shimon Edelman said,

    May 12, 2008 @ 10:56 am

    The way the brain does it causes the child automatically to put words together in certain ways. […] So there you have it—a simple, logical, unified explanation of language acquisition, language evolution, and language creation.

    This reminds me of Steve Pinker's response to Stephen Colbert's challenge to describing how the brain works in five words or less: "brain cells fire in patterns" (see the entire interview here). Pinker's "patterns" surely must include Bickerton's "certain ways" as a special case (and at a more fundamental level of description, too!), so it may appear that Pinker has beaten Bickerton to the explanation. In reality, of course, any brain function must be explained in terms of the computations that support it, and until these are spelled out, the "just so stories" will remain just so. (And if you think that merely calling something a "bioprogram" makes it computationally explicit enough or explanatorily satisfactory, read Drew McDermott's 1976 paper Artificial Intelligence meets natural stupidity).

    Getting back to the topic of this thread, developmental, behavioral, and computational studies suggest that information afforded by social interaction is critically important in learning language (for a summary of the evidence, see this recent review that I wrote with Heidi Waterfall). There are also computational simulations that show how interaction can steer the evolution of language toward the emergence of structure (see for instance Simon Kirby's work). Thus, it would seem that the bird studies that started off this discussion are relevant to language acquisition after all.

    -Shimon

  3. Charles said,

    May 12, 2008 @ 11:23 am

    "If there is any kind of biological program for language (and later I'll suggest the most plausible form this might take) there's no logical reason why it should not kick in for the first generation that needs it."

    While this sentence makes sense in principle, I think they key word is "need". The first generation to grow up hearing a pidgin will need to develop language which is compatible with the previous generation's pidgin, but also sufficient for their own needs – leading to a creole. But in future generations, the "need" for compatibility with the original pidgin drops out. So this could give us at least two generations of very significant language change.

    I figure that we have the capacity, as populations, to advance a language from stage A to B, from stage B to C, but not necessarily from stage A all the way to C in one generation. Our capacity for language will naturally alter certain gaps or defects in what we learn, but only up to a certain threshold; from there on, it's up to the next generation.

  4. Ryan Denzer-King said,

    May 12, 2008 @ 1:46 pm

    Anybody want to attempt an Optimality Theory account of language acquisition? To me what Charles says makes a lot of sense, and also smacks of OT: correspondence constraints which want the second-generation language to match with the first-generation language, and a family of semantic or pragmatic constraints which want language to be maximally expressive.

  5. Derek Bickerton said,

    May 12, 2008 @ 4:30 pm

    Some misapprehensions here! Charles is totally wrong in supposing that the first creole generation "will need to develop language which is compatible with the previous generation's pidgin." In the first place, the pidgin is so variable that you couldn't develop anything "compatible with it" anyway. In the second place, there's no way in which children,. or anyone else, can steer a language in any direction. Language is a process completely below the level of consciousness and ipso facto cannot be consciously controlled. In the third place, there's empirical evidence that it doesn't happen. As I've heard with my own ears (you can read about it all in Bastard Tongues) creole-speaking children repidginize their language when talking to their parents, in other words, rather than tailoring their speech to their parents' needs, they develop a specialized register called "Talking to Pidgin Speakers", just as you or I probably would.

    Then John Roth says, " if the cohort of children has to invent its way through one or more stages it won't have the necessary prerequisites for the next stages when the brain is ready to learn them." Missing the point entirely–the brain isn't learning them. The brain is produciong them. The idea that the brain ios some kind of helpless jelly that can't function unless stuff's punped into it is very general but simply wrong.

    Shimon Edelman blames me for being insufficiently explicit, despite the obvious fact that a single blog posting is hardly the place to lay out a complex theory, so if the five-word version won't cut it, here's the hundred-word version: The brain learns words, knows what they mean, and that verbs aren't complete without their arguments, so it attaches arguments to verbs, working bottom up, and since some verbs (like "tell') need another verb-argument cluster, and since nouns unknown to the receiver can have attached to them another verb-argument cluster that tells you what they are, you get complex sentences for free, and since the verb can remember what got into the structure before what, you get c-command for free (whatever gets attached later to the growing structure c-commands what got attached earlier).

  6. Ryan Denzer-King said,

    May 12, 2008 @ 7:32 pm

    "Language is a process completely below the level of consciousness and ipso facto cannot be consciously controlled."

    Does this mean there aren't subconscious preferences at work? Maybe Charles can clarify, but I wasn't under the impression that he was specifically saying that children consciously attempt to adapt their speech to that of the earlier generation; I thought he was saying that this is an unconscious contraint at work in child language. We change the prefix in- to im- in front of labials, and I think it is missing to point to refrain from saying that people "want" nasals to be homorganic with a following stop. This "want" is not at all conscious, but it still exists as a powerful mechanism for linguistic change.

  7. Mark Young said,

    May 12, 2008 @ 9:47 pm

    I'm not a linguist — I'm just trying to understand the theory, not criticize it!

    So, every child in a social environment learns words for things. Eventually each child will start putting words together to express "bigger" thots. The child makes assumptions about how the words go together because that's what brains do — make assumptions about how things go together.

    In a normal linguistic environment, the child's assumptions will sometimes be "wrong" — according to the pre-existing (and "pre-leveled") assumptions of the adults in that community. The child will notice that their assumptions are wrong and will adjust them.

    But a child raised in a pidgin environment will not get any (or much) such feedback — there is no pre-existing set of assumptions because the local adults are all speaking a pidgin rather than a "proper" language.

    Your (Bickerton's) point, I think, is that the child gets a proper language out of the process anyway. Whatever the child comes up with is not "impoverished" like the pidgin is — it has structures, function words, whatnot. It is not a pidgin; it is not something between a pidgin and a creole; it is a creole.

    Is that a reasonable non-technical statement of the claim? I hope so.

    So if that's right, what about the "multiple generations" question?

    Well, it's not necessarily the case that every child growing up in a particular pidgin environment ends up with the same set of linguistic assumptions — or even a largely overlapping set of assumptions. The children could have languages with overlapping vocabulary, but very different syntax/grammar. So even tho every child speaks a creole language, they speak differing versions of it. In such a case there would still be some "leveling" for the next generation or two to carry out — forging the "dialects" into a language.

    Is it possible that what some people mean when they say that a creole may take several generations to form is that it may take several generations for the creole to "level out" — versus what you mean when you say a creole forms in the first generation (viz, that each child has a fully formed language, despite the possible requirement for future leveling)?

  8. Ewan said,

    May 12, 2008 @ 9:58 pm

    The specifics I was looking for are less about the conception of the language faculty you were trying to present and more about what the big deal about multi-generational creolization is anyway… I agree completely with Mark's comment that

    "The logic of the debate … seems to be understood as follows: if a species-typical behavior appears in the first generation of individuals raised with no conspecific models, then it must be the result of a very specific genetic 'bioprogram'; but if the behavior emerges from development over several generations, it must be the result of general cognitive abilities. Neither of these implications seems to me to be logically valid."

    I was hoping for a reply, and you hinted that (apparently) pidgins and the first generations creoles are worlds apart in terms of their structure and variability – but that could in principle be true of the zebra finches too. It doesn't really seem to be informative. Any further comments?

RSS feed for comments on this post